Kicking around the idea for a DIY polysynth

[uvacom-rotatt]

I'm an EE student, and next year I'll have to do a senior thesis. One thing I've thought about doing is some kind of synthesizer. I think a polysynth would be cool because there aren't very many affordable analog polysynths, and I would consider it a great achievement if I could create something along the level of, say, a Jupiter 8 or an OB-X, that somebody could build on their own from off-the-shelf parts for maybe $1000 (perhaps substantially less but certainly not more). The project would be open source and I may or may not sell the kits myself.

 

My (very preliminary) design constraints:

 

*All analog signal path

*Current production components

*Moderate component cost

*No menus

*Through-hole components only

*Polyphonic operation, most likely 8 voices

*Front panel control of every parameter with no stepping

*Patch memory

 

I don't want to reinvent the wheel, and I don't want to make some kind of

[CfNorENa]

I would consider it a great achievement if I could create something along the level of, say, a Jupiter 8 or an OB-X, that somebody could build on their own from off-the-shelf parts for maybe $1000 (perhaps substantially less but certainly not more)

 

That would be a great achievement indeed! Probably fair to say that you would have the entire HC KSS community pulling for you to succeed

[ElectricPuppy]

[SoundwaveLove]

Seems like a big project. You may want to start a bit simpler

 

8 voices is a lot. With analog synths the sounds can be so thick you don't really need more then 2, chords don't sound good with FAT sounds. I'd be happy with 2. Especially if you could split the keyboard and route them through separate filters. One multi-mode, and one low pass.

 

if you have two oscillators, then how about just one LFO, and one of the oscillators can be switched to LFO like on the minimoog.

 

Also if you have two envelopes, you can have a loop option so they can be used like LFO's too, but better.

 

Also you need be be able to midi sync the LFO's and envelopes.

[soundwave106]

I would consider it a great achievement if I could create something along the level of, say, a Jupiter 8 or an OB-X, that somebody could build on their own from off-the-shelf parts for maybe $1000 (perhaps substantially less but certainly not more).

 

 

I am not an EE, just a computer programmer, but I'm speaking off the cuff.

 

But the off-the-shelf parts thing might be your biggest humdinger. From what I see, "off the shelf", hand-solderable parts for an analog polysynth will probably turn into monsters like the 4 Voice or the CS-80, with somewhat more limited digital control. (Hence why many polys combine circuits on a chip.)

 

If you have access to modern assembly shops, you can make something like the Omega 8s and Jomox Sunsyns. Which I understand is a discrete signal path, but still probably uses surface mount components to reduce size. My impression was that hand-soldering surface mount is a PITA, reducing the usefulness of a kit idea. Maybe I'm wrong.

 

But hey -- as an EE, I'm sure you know, or at least can easier research, something I don't know. Wish you luck!

[grumphh]

I want to build a really good-sounding, usable polysynth. Spec-wise, I imagine a voice architecture like this: *Osc 1 - saw, tri, pulse, sub - all simultaneously available and mixable *Osc 2 - saw, tri, pulse - single output with level control *VCF - state variable type with outputs for LPF, BPF, and HPF - 12/24dB switch for LPF *VCA - just the basics, emphasis on low noise and very fast operation *2 ADSR envelopes hardwired to VCA and VCF cutoff, and a small handful of other parameters assignable *2 LFOs, one hardwired to pitch/vca, the other to VCF, again with a few other assignable destination More to come...

Take away the BP and HP filter modes and one LFO, add a couple of modulation options at the VCO section and you will have yourself a Jupiter 8.

 

A worthy project

 

 

Yes yes, i know the JP8 has a HP filter - but i would imagine the idea of this one is that it should be a "proper" resonant filter...

[ChristianRock]

Yeah I'd start slowly and build a monophonic synth first, then I'd go from there to the poly... pretty much everybody did this, except for Dave Smith I think...

[evildragon]

JP8 has a highpass, doesn't it?

[ElectricPuppy]

You're gonna need an MCU and some firmware along with some DAC's and maybe some CV multiplexing, unless you want a fully-manual, non-MIDI mono synth.

 

Design it so that each voice is its own card; that way a builder can start with one voice and build more as their budget allows.

 

$1000 might be optimistic. Little individual parts like resistors and such are cheap, yes, but they're not what racks-up the cost of synths like this. It's the pots, knobs, connectors, keyboard, and all that other hardware that adds up fast.

[grumphh]

JP8 has a highpass, doesn't it?

A useless one

It is more like a bass EQ on a home stereo except that you can't boost it. All it does is remove bottom end.

 

...and now learn to read the small print too...

 

[evildragon]

Damn, never really saw the small print because I was reading too fast!

 

You got me there, grumphh! :poke:

 

 

Removing bottom end could be useful for some sounds, actually.

[grumphh]

Damn, never really saw the small print because I was reading too fast! You got me there, grumphh! :poke:

 

Maybe i should consider a career in the insurance business?

[evildragon]

Hahaha

 

Well, I'm tired as it is after today's college exam, so don't mind me. I bet others catched the small print.

[Yoozer]

 

My (very preliminary) design constraints: *All analog signal path *Current production components *Moderate component cost *No menus *Through-hole components only

 

 

Start with this. Do it the Studio Electronics way - make voice boards. That way you can build your own 1, 2, 4, 5, 8-voice polyphonic synth while you can spend the rest of the time worrying about voice assignment and -stealing.

 

 

Spec-wise, I imagine a voice architecture like this: *stuff*

 

Sounds good. Where's the noise waveform? Ringmod and sync are expected, too - otherwise you've just made an improved Poly-61.

[evildragon]

Why not be brave and add a 3rd envelope with several modulation destinations? Also, don't hardwire LFOs, make them have 4-8 destinations each, would be more flexible

 

Also, you could start with a monophonic voice board, then just add them up and worry about voice assignment.

 

Several other suggestions:

 

* make oscillators the same, with same options. Add sync and FM to second one. Give subosc to each osc.

* white/pink noise generator in mixer section. Ditto ringmod.

* about filter, why not implementing 2 2-pole SVF in series, each having their Mode sweep knob (so you can sweep between LP-BP-HP), and between them you have a Separation knob to differentiate cutoff points? This way you could get a really flexible 4 pole filter with wicked modulations (mode sweep should be modulatable, of course )

[AnCap]

It would seem to me that making a cool, possibly viable as a business venture, mono would be easier. There will always be people wanting a cool moderately priced mono. A two vco mono with some unique features would sell if priced decently and made well. I am not sure such a product even exists.

[mate_stubb]

I am an EE, and I have built a polysynth.

 

You increase the level of complexity an order of magnitude by adding patch memory.

 

You will spend more than $1000. Way more, unless you etch your own boards.

[bruto]

There is a second way to go poly - a polyphonic CV keyboard like the Roland M-184. It's four note poly and can control 4 monophonic modules via 4 CV and Gate outputs.

 

The M-184s are getting stupidly expensive and seem to go up every year. The last one I saw on ebay sold for over $1500. Somewhere I read of someone building a clone - although I can't remember where I read it.

 

Anyway, if you had an M-184, you could run it to four single or double oscillator monosynths and have a sweet poly similar to an Oberheim 4 voice.

[The Real MC]

 

I am an EE, and I have built a polysynth. You increase the level of complexity an order of magnitude by adding patch memory. You will spend more than $1000. Way more, unless you etch your own boards.

 

 

I agree - polysynths aren't just an EE challenge, they are a systems design challenge.

 

There is physics of components. Monosynths are one thing, but polysynths are another. Cramming sixteen heat sensitive VCOs in a compact case presents a challenge in tuning stability that has confounded designers of the early polysynths (rev 1 P5, memorymoog, Oberheim, roland, yamaha). Ultimately it came down to taming the physics of sensitive components and proper dissipation of heat.

 

PC board design of a high frequency microcontroller system is another field of study. Don't even attempt this one without proper bench test instruments like logic analyzers, in circuit emulators, or wide bandwidth oscilloscopes.

 

I don't mean to discourage but you should be aware what you are getting into.

 

I recommend studying the Memorymoog, rev 3 Prophet 5, or Rhodes Chroma service manual. Not just schematics but the SERVICE MANUAL, read the text.

[ChetSinger]

I've spent years building synth electronics as a hobby. I started with a CMOS phase-locked loop and ended up designing with DSP chips. My 2 cents include:

 

 

[mate_stubb]

 

I recommend studying the Memorymoog, rev 3 Prophet 5, or Rhodes Chroma service manual. Not just schematics but the SERVICE MANUAL, read the text.

 

 

+1.

 

In fact, I studied the Prophet 5 schematics for months. They feature discrete TTL gates, and so the design is something that is valuable to learn from.

 

Chroma is considerably more advanced.

[ElectricPuppy]

I've looked over the Chroma schematics. The charge-pump VCO in particular is interesting, I don't know of any other commercially-built analogs that use that.

 

I've toyed with the idea of remaking my own Chroma, especially since there's that whole new computer somebody had designed to replace the original.

[mate_stubb]

Puppy haz a Chroma!? Must fix!

 

One cool thing about that discrete VCO design is that you can vary the sawtooth symmetry (like PWM) and that in the middle it turns into a triangle 1 octave higher.

[ElectricPuppy]

No, no, I don't haz a Chroma, much as I would love one. I meant, I'd like to build a new one. There was a shell of one on Ebay a few weeks ago; complete case and keyboard, but no electronics... I seriously considered it.

[uvacom-rotatt]

Whoa! Glad to see there is a lot of interest in this. I ran out of time this afternoon before I could fully describe the system.

 

Okay, first off, I absolutely agree that the first step is getting a monophonic synthesizer working. That is, in fact, basically what I plan on doing.

 

Second, I am aware of the challenges in heat, packaging, and cost. There are a few things I hadn't mentioned that mitigate this. One of the things that spawned this idea for me is that I've seen arduinos that can do simple DSP. I'm already familiar with the Atmel AVR and I'm learning a great deal about FPGAs so I feel like the technology available can give great performance for things like providing the modulation signals (ADSRs, LFOs, keytracking, etc.) and the clock pulses for the oscillators (the waveshaping, etc. would all be analog still).

 

I'd like to use OTAs and opamps where ever possible because that saves space and reduces complexity. But only where the sound is not severely compromised. For example, I'm really concerned with the performance of the VCA. It has to have low noise, fast slew, and a good dynamic range.

 

Obviously I'm not the first person to come up with the idea of using digital control to build a hybrid device. But I've done some work in modeling analog systems with Morgana which I learned a lot from, so I think I would enjoy the challenge of developing say, a pitch drift algorithm for the oscillators that sounds indistinguishable from a full VCO but gives overall good pitch stability. Some people have said it's as easy as running a noise source through a lowpass filter, sampling it and using it as a seed, and that certainly sounds like a good start. But I want to do some real research - build a good VCO, track it's pitch, and study the frequency spectra of its drift. Similarly, it is important to me to spec a processor/controller platform with enough grunt to deliver very fast, high resolution modulation signals that do not deliver appreciable quantization and have very natural slopes.

 

This also makes the big challenge of designing a system which has totally analog front panel control for things like pitch, filter cutoff, etc. but also patch memory easier to swallow, because far fewer analog parameters need to be sampled, stored, and recalled. Anything related to pitch or modulation will be almost trivial to implement.

 

So I think the project is a challenge. I'm thinking ahead and trying to tackle the problems that I understand now. It will be about a year and a half before I'm done with my thesis, and I think if I have a semi-reliable one voice prototype without patch memory I'll be satisfied. I think it will be a few years before people can buy a kit and built their own polysynth.